Business Cycle Tracer revisited ECFIN Workshop, november 2012 Erik Slentoe Statistics Denmark

Composite Confidence Indicator Traditional Composite Confidence Indicator for industry Boom and recession periods shows during time Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

A tool to analyze and illustrate economic cycles in business Read counter clock wise Light blue 2005 – 2008 - dark blue 2009 to present 4 quadrants: above/below trend and increasing/decreasing Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

Construction the Business Cycle Tracer is a 2-step procedure: Generating a indicator for the graph Construction the graph First about the construction the graph Follows the method used by ECFIN version developed by Christian Gayer about 2007 (?). Based on method developed by IFO, Germany. Also used by e.g CBS, OECD and others. 4 12/31/2018 12/31/2018

Construction principle: x Konjunkturindikator konjunkturværdi Business Cycle Tracer Confidence Indicator Indicator value mth-to-mth changes y y-value is transferred directly x-value is increase/decrease in y values: yt – yt-1 Dots on curve marks months 5 12/31/2018 12/31/2018

Data is smoothed with Hodrick Prescott filter (lambda 69) (minimizing problem: fit to data vs smooth data) Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

Turning point problem May 2009 Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

May 2009 Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

The turning point issue is a draw back and limit the use of the graph. Developing a method to minimize revision-surprises is desirable. Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

Sensitivity illustration helps ”forecast” reading Dotted lines indicates development if x-value of present month was either plus or minus std.dev. of the x-time series. Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

Contructing the Business Cycle Tracer is a 2-step procedure: Generating a indicator for the graph Construction the graph Now about generating the underlying indicator for the graph 11 12/31/2018 12/31/2018

Instead of Composite Confidence Indicator is generated a “confidence value” based on Principal Component Analysis Composite confidence indcator Confidence value (PCA) Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

By PCA-analysis is generated one new variable from a number of variables Original variables (x, y) highly correlated New variables (p, q) Transforming data into a new coordinate system (p,q) p accounts for most of the variance in data q accounts for almost nothing – can therefore be discharged Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

Method works because the basic indicators normally are highly correlated. However this is not the case for Denmark regarding the Industry sector (manufacturing) Production, developed Export order books Production, expectations Overall order books Stock of finished products Stock of finished products is omitted from the new variable Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

Original variables (x, y) not correlated New variables (p, q) Both p and q accounts for the variance in data Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

Principalkomponentanalyse (PCA): KBI Variabel PC1-loadings Production, developed 0,93 Overall order books 0,95 Export order books Stock of finished products -0,08 Production, expectations 0,80   Explanation (of total varians) 66% Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

PCA generated variable seems a statistical better ”composite indicator”, than tradition Composite Confidence Indicator (simple average), however, as shown, it has its pitfalls. The information if the basis-indicator ”Stock of products” is omitted, although it might be the first indicator to notice about changes. Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

All in all an interesting and visual appealing graph Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012

end References Die ifo Konjunkturuhr: Zirkulare Korrelation mit dem Bruttoinlandsprodukt http://www.cesifo-group.de/pls/guest/download/ifo%20Schnelldienst/ifo%20Schnelldienst%202010/ifosd_2010_5_4.pdf   Gayer, Christian (2007). The Economic Climate Tracer A tool to visualise the cyclical stance of the economy using survey data http://www.oecd.org/dataoecd/12/47/39578745.pdf van Ruth, Schouten and Wekker (2005) : The Statistics Netherlands’ Business Cycle Tracer. Methodological aspects; concept, cycle computation and indicator selection http://www.cbs.nl/NR/rdonlyres/253FD272-B93E-46FF-A474-1E5A396C81F1/0/2005methodebusinesscycletracerart.pdf Erik Slentoe, Statistics Denmark, ECFIN Workshop, Bruxelles Nov. 2012